Content based image retrieval with user relevant feedback

Abstract

Content Based Image Retrieval (CBIR) is used to retrieve the exact meaningful images from the Image Database using image matching and retrieval based on semantic features like color, texture, shape and integrated odels. The research work explores semantic based image retrieval system with feedback using various features extracted from the images.In this work, CBIR methodology is experimented by exploiting the semantic features such as Coarseness, Contrast, Directionality, Local Binary Pattern, Local Tetra Pattern along with Relevance Feedback mechanism to retrieve the user expected result. Two similar images that are semantically different are identified using the selected features. To speed up the feature based image retrieval process, Binary Wavelet Transform is modified and proposed as new feature. The extracted features of the query image are represented as a feature vector in the form of one dimensional column vector. The same step will be repeated for all images presented in the database. Then the Euclidean distance calculation is applied between the query image feature vector and the database image feature vector to calculate the similarity distance. Based on the similarity distance, a set of most similar images corresponding to the query image is retrieved. It can be viewed by the user and relevant feedback is given to identify the user desired images using Graphical User Interface (GUI) application. The proposed method produced accuracy between 98% and 100% for the Corel database using interactive user feedback system. Similarly, based on Average Retrieval Rate the proposed system produced 98.41% for Vistex, 96.2% for Brodatz and 95.8% for Alot databases. For color image retrieval application, the proposed scheme can be treated as an aggressive example. newline